file.cobra(11): error: twice in error messages.

[nevdelap]

Hi Chuck. I searched for "error twice", couldn't think of anything, maybe it's already been touched on...

I wonder why the duplication of filename, line number and "error:" in error messages that want to tell us the column?

test.cobra(11): error: test.cobra(11,1): error: Cannot mix tabs and...

rather than...

test.cobra(11,1): error: Cannot mix tabs and...

[Charles]

Looks like a bug. Other errors I tried don't have this problem, so it may only be happening for errors generated by the parser. Another ticket, please.

[nevdelap]

Done. I'd thought it might have something to do with tools reading the compiler output so I wanted to check before ticketing. #212.

[hopscc]

patch for this on the ticket (ticket:212)

[Charles]

This bug is now fixed and ticket is closed.

[hopscc]

The fix dropped display of column number an error was detected on (where available) - probably an unintended regression
nevdelap discovered and reported against ticket.

I reopened the ticket and posted a patch that continues to provide the column info if available.

[Charles]

I vaguely recall not reporting column info in Cobra errors early on because Cobra wasn't counting out the tabs, so the column didn't really match what I saw in my editor. Maybe we added that at some point and I've forgotten. Are those column numbers correct in your editor?

[hopscc]

Yeah pretty much.

[Charles]

When I try an error from one of the binding phases like "x = badname", I get:

foo.cobra(5): error: Cannot find "badname".

I'm now more certain that column info was left out because it does match editors. When I'm three tabs over, my editors tell me I'm on column 13, not 4 which is what Cobra would tell me.

[hopscc]

Presumably you mean 'does not match editors'.

Cobra cant ever be correct for this (without talking to your editor) because it depends on what your tab expansion is set to in your editor
- at best if you want to try and virtualise this you assume a tab expansion of 4 (which is what cobra seems to equate with spaces ) and accumulate a Tab count and multiply the number of Tabs by that
- fine if it is correct to the editor settings really confusing if not. And I'm sure I'm not the only one who just loves it when a compiler lies to you.....

On the other hand you could do what every compiler has done since the first one was carved from primordial mud and just quote the character count offset from the start of the line (4 in your example)
and assume the user/developer can work it out.

Most serious programming editors have a mode or command that will give you this on a line (char offset vs display offset) -
Its just one of the things you have to know how to handle if you use/allow Tabs in source code...

Its still needed because it gives the user at least an inkling to exactly where on a line the compiler thinks it detected the problem which may or may not be obvious from the rest of the error message....